[cifs-discuss] Potential performance issue with CIFS server snv_95

[Daniel Liebster]

Hello,

Question about some rather unexpected performance difficulties..

One of our genomics staff wrote a python script to test out CIFS performance, and it seems to have found some performance problems with out CIFS server...

In the attached screenshot (from the Windows 2003 box that the script is running on) the first , irregular part of the traffic flow is the python script's random access tests to a windows 2003 server, and then a HW based CIFS fileserver.
The high flat lines are the same test going to the 1GB and and then 10GB interface of the x4500 CIFS server. The same python script really dogs the CIFS server for some reason...( the test takes bout twice as long to complete on the Sun as either the Windows 2003 server or the HW based file server....)


In terms of regular file copy performance, robocopy reports more expected results. In a test copy of a 47 file, 40GB directory from the test windows server, the results were..

To windows 2003 server:      1041.682 MegaBytes/min.
To openSolaris CIFS-server:    1770.794 MegaBytes/min.
To HW filer (FC):       1824.360 MegaBytes/min. 

Any Ideas on why the python script shows such odd/slow network performance?
--
This message posted from opensolaris.org